Ancestry prediction in Singapore population samples using the Illumina ForenSeq kit. (November 2017)

• Record Type: Journal Article
• Authors: Ramani, Anantharaman; Wong, Yongxun; Tan, Si Zhen; Shue, Bing Hong; Syn, Christopher
• Is Part Of: Forensic science international. Volume 31(2017)
• Page Start: 171
• Page End: 179
• Publication Date: 2017-11
• Subjects: Bio-geographic ancestry prediction -- Massively parallel sequencing -- Singapore population -- ForenSeq -- Snipper -- STRUCTURE; Forensic genetics -- Periodicals; Génétique légale -- Périodiques; Forensic genetics; Electronic journals; Periodicals; 614.1
• DOI: 10.1016/j.fsigen.2017.08.013
• Languages: English
• ISSNs: 1872-4973

Abstract

Highlights: Ancestry prediction capabilities of ForenSeq kit evaluated in 1030 Singapore samples. 59 ancestry SNPs and phenotypic SNPs with AIM properties shortlisted. Ancestry modelling using Snipper and STRUCTURE with separate training and testing sets. Correct ancestry prediction in 91%–94% of Chinese, 80%–88% of Malay and 91% of Indian individuals. Abstract: The ability to predict bio-geographic ancestry can be valuable to generate investigative leads towards solving crimes. Ancestry informative marker (AIM) sets include large numbers of SNPs to predict an ancestral population. Massively parallel sequencing has enabled forensic laboratories to genotype a large number of such markers in a single assay. Illumina's ForenSeq DNA Signature Kit includes the ancestry informative SNPs reported by Kidd et al. In this study, the ancestry prediction capabilities of the ForenSeq kit through sequencing on the MiSeq FGx were evaluated in 1030 unrelated Singapore population samples of Chinese, Malay and Indian origin. A total of 59 ancestry SNPs and phenotypic SNPs with AIM properties were selected. The bio-geographic ancestry of the 1030 samples, as predicted by Illumina's ForenSeq Universal Analysis Software (UAS), was determined. 712 of the genotyped samples were used as a training sample set for the generation of an ancestry prediction model using STRUCTURE and Snipper. The performance of the prediction model was tested by both methods with the remaining 318 samples. Ancestry prediction in UAS was able to correctly classify the Singapore Chinese as part of the East Asian cluster, while Indians clustered with Ad-mixed Americans and Malays clustered in-between these two reference populations. Principal component analyses showed that the 59 SNPs were only able to account for 26% of the variation between the Singapore sub-populations. Their discriminatory potential was also found to be lower (GST = 0.085) than that reported in ALFRED (FST = 0.357). The Snipper algorithm was able to correctly predict bio-geographic ancestry in 91% of Chinese and Indian, and 88% of Malay individuals, while the success rates for the STRUCTURE algorithm were 94% in Chinese, 80% in Malay, and 91% in Indian individuals. Both these algorithms were able to provide admixture proportions when present. Ancestry prediction accuracy (in terms of likelihood ratio) was generally high in the absence of admixture. Misclassification occurred in admixed individuals, who were likely offspring of inter-ethnic marriages, and hence whose self-reported bio-geographic ancestries were dependent on that of their fathers, and in individuals of minority sub-populations with inter-ethnic beliefs. The ancestry prediction capabilities of the 59 SNPs on the ForenSeq kit were reasonably effective in differentiating the Singapore Chinese, Malay and Indian sub-populations, and will be of use for investigative purposes. However, there is potential for more accurate prediction through the evaluation of other AIM sets.